第 6 章 内核

在 Red Hat Enterprise Linux 7.1 内核中添加了 libceph.ko 和 rbd.ko 模块。这些 RBD 内核模块允许 Linux 主机将 Ceph 块设备视为常规磁盘设备条目，可以挂在到某个目录并使用标准文件系统格式化，比如 XFS 或者 ext4。

注：Red Hat Enterprise Linux 7.1 目前不支持 CephFs 模块 ceph.ko。

在 IBM System z 架构中的 Red Hat Enterprise Linux 7.1 启用微代码级升级（MCL）。可在不影响 flash 存储介质 I/O 操作的情况下应用这些升级，并通知用户更改了 flash 硬件服务等级。

Red Hat Enterprise Linux 7.1 introduces kpatch, a dynamic "kernel patching utility", as a Technology Preview. The kpatch utility allows users to manage a collection of binary kernel patches which can be used to dynamically patch the kernel without rebooting. Note that kpatch is supported to run only on AMD64 and Intel 64 architectures.

Red Hat Enterprise Linux 7.1 启用在多个 CPU 中引导 crashnernel 的功能。这个功能作为技术预览支持。

Red Hat Enterprise Linux 7.1 引进了 dm-era 设备映射器目标作为技术预览。dm-era 可跟踪在用户定义的时间段（即“era”）内写入的块。每个 era 目标事务将当前 era 作为单调增长的 32 位计数器维护。这个目标允许备份软件，以便追踪自上次备份后产生变化的块。它还可启用缓存的部分失效内容，以便在返回经销商快照后恢复缓存一致性。dm-era 目标主要与 dm-cache 目标配对。

The Cisco VIC Infiniband kernel driver has been added to Red Hat Enterprise Linux 7.1 as a Technology Preview. This driver allows the use of Remote Directory Memory Access (RDMA)-like semantics on proprietary Cisco architectures.

The paravirtualized hardware RNG (hwrng) support for Linux guests via virtio-rng has been enhanced in Red Hat Enterprise Linux 7.1. Previously, the rngd daemon needed to be started inside the guest and directed to the guest kernel's entropy pool. Since Red Hat Enterprise Linux 7.1, the manual step has been removed. A new khwrngd thread fetches entropy from the virtio-rng device if the guest entropy falls below a specific level. Making this process transparent helps all Red Hat Enterprise Linux guests in utilizing the improved security benefits of having the paravirtualized hardware RNG provided by KVM hosts.

之前，调度程序负载平衡代码在所有闲置 CPU 之间平衡负载。在 Red Hat Enterprise Linux 7.1 中，只有需要对 CPU 进行负载平衡时才会代表闲置 CPU 执行闲置负载平衡。这个新行为降低了非闲置 CPU 的负载平衡率，进而减少了该调度程序的非必要任务负载，从而提高其性能。

修改了该调度程序的行为，如果有可运行的任务，则会停止在 newidle 平衡代码中搜索任务，从而提高性能。

Red Hat Enterprise Linux 7.1 添加在运行时支持超大页面分配的功能，可让 1GB hugetlbfs 用户指定在运行时为哪个节点分配非一致内存访问（NUMA）节点。

Red Hat Enterprise Linux 7.1 引进了新的锁定机制，MCS 锁。这个新锁定机制极大减少大系统中的 spinlock 负担，让 spinlock 在 Red Hat Enterprise Linux 7.1 中更有效地运行。

Since Red Hat Enterprise Linux 7.1, the kernel process stack size has been increased from 8KB to 16KB to help large processes that use stack space.

In Red Hat Enterprise Linux 7.1, the uprobe and uretprobe features work correctly with the perf command and the systemtap script.

End-To-End data consistency checking on IBM System z is fully supported in Red Hat Enterprise Linux 7.1. This enhances data integrity and more effectively prevents data corruption as well as data loss.

In Red Hat Enterprise Linux 7.1, the deterministic random bit generator (DRBG) has been updated to work on 32-bit systems.

As a Technology Preview, the NFSoRDMA service has been enabled for Red Hat Enterprise Linux 7.1. This makes the svcrdma module available for users who intend to use Remote Direct Memory Access (RDMA) transport with the Red Hat Enterprise Linux 7 NFS server.

The Kdump kernel crash dumping mechanism on systems with large memory, that is up to the Red Hat Enterprise Linux 7.1 maximum memory supported limit of 6TB, has become fully supported in Red Hat Enterprise Linux 7.1.

With Red Hat Enterprise Linux 7.1, the Kdump crash dumping mechanism is supported on machines with enabled Secure Boot.

Red Hat Enterprise Linux 7.1 introduces support for firmware-assisted dump (fadump), which provides an alternative crash dumping tool to kdump. The firmware-assisted feature provides a mechanism to release the reserved dump memory for general use once the crash dump is saved to the disk. This avoids the need to reboot the system after performing the dump, and thus reduces the system downtime. In addition, fadump uses of the kdump infrastructure already present in the user space, and works seamlessly with the existing kdump init scripts.

As a Technology Preview, support for the Runtime Instrumentation feature has been added for Red Hat Enterprise Linux 7.1 on IBM System z. Runtime Instrumentation enables advanced analysis and execution for a number of user-space applications available with the IBM zEnterprise EC12 system.

Cisco Unified Communication Manager (UCM) servers have an optional feature to provide a Cisco proprietary User Space Network Interface Controller (usNIC), which allows performing Remote Direct Memory Access (RDMA)-like operations for user-space applications. As a Technology Preview, Red Hat Enterprise Linux 7.1 includes the libusnic_verbs driver, which makes it possible to use usNIC devices via standard InfiniBand RDMA programming based on the Verbs API.

The i40e and i40evf kernel drivers have been updated to their latest upstream versions. These updated drivers are included as a Technology Preview in Red Hat Enterprise Linux 7.1.